The use of Boosted Decision Trees method for jet classification on the first level trigger (run 2)"

Postgraduate Thesis uoadl:1320765 566 Read counter

Unit:
Κατεύθυνση Πυρηνική Φυσική και Φυσική Στοιχειωδών Σωματιδίων (ΒΑΣΙΚΗ ΦΥΣΙΚΗ)
Library of the School of Science
Deposit date:
2015-10-12
Year:
2015
Author:
Καραθανάσης Γεώργιος
Supervisors info:
Παρασκευάς Σφήκας Καθηγητής
Original Title:
Χρήση της μεθόδου των Boosted Decision Trees στην αναβάθμιση του σκανδαλιστή πρώτης τάξης (run 2)
Languages:
Greek
Translated title:
The use of Boosted Decision Trees method for jet classification on the first level trigger (run 2)"
Summary:
During the LS1, LHC has been improved, making possible to reach energies up to
6.5TeV per beam. This has increased the number of pile-up interactions at 50 per
bunch crossing. Thus the trigger rate of the first level trigger (L1) is being
affected. Given that L1 can not excess the limit of 100kHz, we had to make it
more
efficient. The improvements included new electronic cards and the introduction
of a
novel trigger architecture. With those changes, we can develop better and more
efficient algorithms. In this paper we have used the Boosted Decision Trees(BDT)
method to discriminate jets originating from pile-up events, from signal. BDT
is a
supervised learning algorithm, thus requires a sample of training events. After
some
tuning we took the optimal set of options. Finally we have compared the
efficiency
of the previous and BDT method, which proves that BDT is more efficient.
Keywords:
CERN , CMS, First Level Trigger, BDT
Index:
No
Number of index pages:
0
Contains images:
Yes
Number of references:
46
Number of pages:
86
document.pdf (3 MB) Open in new window